Replace the `eos_token` with `<eot_id>`, to prevent model from going on unnecessarily long

#2
No description provided.

Thank you for adding this model, and for flagging the unnecessarily long completions issue!

This was an issue with the original Llama3 tokenizer. I have just updated this in the sqlcoder repo, too.

rishdotblog changed pull request status to closed
rishdotblog changed pull request status to open
abhishek-ch changed pull request status to merged

Sign up or log in to comment